基于ARIMA的传染病发病率趋势预测模型实证研究

Trend prediction modeling empirical research of infectious disease morbidity based on ARIMA

目的探讨ARIMA模型设计流程和调整优化问题,对部分甲乙类传染病发病率进行预测。方法根据国家公布的1990-2000年的痢疾、伤寒、布鲁菌病和流行乙脑时序数据,作纯随机性和平稳性检验,低阶差分平稳序列绘制自相关和偏自相关图,借助SAS软件实现模型定阶调试、参数识别、残差检验、模型拟合预测。结果发病率数据存在短期相关性,一阶差分提取趋势信息,经BIC准则定阶疏系数模型,逐步经验调试与优化,参数显著性和残差白噪声检验有效,模型预测有实用价值。结论 ARIMA法对于传染病发病率等卫生预测问题有适用性、代表性,建模设计应针对实际资料进行综合研究。

Objective To discuss design process and adjustment and optimization problem of ARIMA model,and forecast some class a and b infectious diseases morbidity. Methods The time series data from 1990 to 2000 of dysentery,typhoid fever,brucellosis and Japanese encephalitis were used to randomness and stationary test,after low order difference,smooth and steady sequence were used to make autocorrelation and partial autocorrelation chart. SAS was used to complete order determination,debugging,parameter identification and residual examination optimization. Results Morbidity existed short-term correlation,first difference could extract tendency information,BIC criterion could determine dispersed coefficient model,debugging and optimization was done by experience,parameter significance and white noise examination of residual was valid,model prediction had practical value. Conclusion ARIMA method had applicability and typical to infectious disease morbidity and other hygiene forecast problems,modeling project should given the comprehensive research according to practical data.